How accurate is phone timing?

What we claim today

Nothing, numerically. We have not yet run StatLap beside a certified system and published the results, so we do not quote an accuracy figure, a margin, or a comparison. Anyone who quotes one without the trials behind it is guessing, and you should hold us to the same rule.

What we do say is structural, and you can check every line of it yourself: the finish is read from the video frame the athlete is crossing in, so the resolution of a time is the length of one frame; the clock starts from the gun rather than from a thumb; and every result page is marked app-timed, not certified.

When we have trials, they go on this page in full, including the runs that went badly. If you want to be one of them, the section at the bottom says how.

What the phone actually measures

The start

The start phone listens for the gun and marks the moment it hears it. It learns your starter's sound first, so a whistle, a clap or a pistol all work. That moment, not a person's thumb, is time zero for every phone on the code.

The finish

The finish phone records the line and looks for motion in each lane's band. The time it reports is the frame in which the athlete is crossing, which is why the frame rate matters more than anything else on this page.

The gap between phones

Neither phone trusts its own clock. Each one keeps asking the timing server what time it is, times the round trip, keeps the fastest exchanges and takes the middle of them, which is how computers have synchronised clocks for forty years. Both phones then work in that shared clock, so the gun's moment means the same thing at the finish line.

How well they agree, on screen

You do not have to take that on trust. Both phones show Clock sync and a figure in milliseconds, which is how closely they currently agree. It turns amber above 30 milliseconds, which on a track usually means the network, not the phones. Check it before you trust a time, the way you would check a tape measure.

Speed gates

The same motion detection at two points a known distance apart. The distance you type is part of the measurement: type it wrong and the speed is wrong by the same fraction, however good the timing is.

Where the error comes from, worst first

Frame rate

A time cannot be finer than one frame. At 30 frames a second a frame is 33 milliseconds, at 60 it is 17, at 120 it is 8. Film the finish at the highest rate the phone offers in good light, and this is the smallest it can be.

Where the camera sits

The phone must be square to the finish line, so the line runs straight across the frame. Aimed down the track at an angle, a runner in the far lane appears to cross early or late, and no software fixes it. Square to the line, hip height, on something that does not move.

The network between the phones

The shared clock is only as steady as the connection it is measured over. A quiet hotspot holds it tight; a school Wi-Fi full of phones at a meet makes it wander, and an error there moves every time on the day by the same amount, which is the kind you do not spot by reading the results. This is the one source of error the app puts a number on, so watch the Clock sync figure.

Sound reaching the start phone

Sound covers about a metre every three milliseconds. The phone must be near the starter, not near the finish, or every time on the day is early by the travel time.

Lane bands

The finish phone times whatever crosses a lane's band. A band drawn too wide catches the athlete in the next lane; too narrow and it misses a lean. This is the most common cause of a time that looks impossible.

What the athlete leans with

The detector fires on motion crossing the line, not on a torso specifically. An arm or a leading knee can cross first. Against a certified system, which reads the torso, that is a real and consistent difference.

What it is not

StatLap is not a certified fully automatic timing system, and its times are not eligible where a certified mark is required. Every results page says app-timed. It is built for practice, time trials, dual meets, rankings and team records, which is most of what most programmes do most weeks. If a mark has to be certified on the day, use a certified system.

It is also not hand timing. A thumb on a watch carries a person's reaction time at both ends, and governing bodies add a conversion factor to hand times for exactly that reason. Starting from the gun removes the reaction at the start; the camera removes the guess at the finish.

Check it yourself in twenty minutes

1. Set the finish phone square to the line, hip height, and film at the highest frame rate it offers. Put the start phone with the starter.
2. Run six trials of the same thing, a 60 or a 100, with whatever you normally use running at the same time: a certified system if your meet has one, another app, or two people on watches.
3. Write down both times for every trial, including the ones where something went wrong. The runs that fail are the useful ones.
4. Note the conditions: frame rate, where the phones were, the network, and the light. A result with no conditions attached cannot be repeated.
5. Send it to support@statlap.com. We will publish it here as it came in, and if it shows StatLap is off, that goes here too.

Why we are asking rather than telling

A number we generated ourselves in a car park proves nothing to a coach with a season on the line. Trials from real tracks, published whole, are worth more than a figure in a marketing line, and they are the only kind we will put on this page. Write to support@statlap.com.